Real-world handheld smartphone videos for nystagmus detection in acute vertigo: a feasibility study

Background Accurate identification of nystagmus is fundamental to the assessment of acute vertigo but remains challenging in emergency department (ED) settings. Although smartphone-based approaches have been proposed as scalable solutions, their feasibility under real-world acquisition conditions remains uncertain. Objective This study aimed to evaluate the feasibility of detecting nystagmus from handheld smartphone videos acquired in a real-world ED. Methods A subset of 67 videos from a published handheld smartphone-based acute vertigo dataset was analysed. A salient feature-based eye-tracking algorithm tracked stable ocular and periocular (facial) landmarks across frames. Horizontal displacement signals were extracted, and nystagmus episodes were identified through signal analysis and video verification. Results Consistent eye-movement signals were successfully extracted from handheld smartphone recordings acquired during the acute phase of vertigo despite real-world challenges, including variable illumination, camera motion, image instability, and patient-related factors. Nystagmus patterns observed on video were accurately reflected in the corresponding signal traces, including very low-amplitude nystagmus that was barely discernible to the naked eye. Conclusions Our findings demonstrate that handheld image acquisition using standard, low-cost smartphones can capture clinically meaningful ocular motor information in real-world emergency care settings. This proof of principle lays the groundwork for the development of more ecologically valid scalable, telehealth-compatible tools for acute vertigo assessment. Further studies are needed to automate detection and validate performance in larger prospective cohorts.